Enhancing Rover Mobility Monitoring: Autoencoder-driven Anomaly Detection for the Mars Science Laboratory

Over eleven years into its mission, the Mars ScienceLaboratory remains vital to NASA’s Mars exploration. Safe-guarding the rover’s long-term functionality is a top mission pri-ority. In this study, we introduce and test undercomplete autoen-coder models for detecting drive anomalies, using telemetry datafrom wheel actuators, the Rover Inertial Measurement Unit(RIMU), and the suspension system. Our approach enhancespost-drive data analysis during tactical downlink sessions. Weexplore various model architectures and input features to un-derstand their impact on performance. Evaluating the mod-els involves testing them on unseen data to mimic real-worldscenarios. Our experiments demonstrate the undercompleteautoencoder model’s effectiveness in detecting drive anomalieswithin the Curiosity rover dataset. Remarkably, the modeleven identifies subtle anomalous telemetry patterns missed byhuman operators. Additionally, we provide insights into optimaldesign choices by comparing different model architectures andinput features. This research significantly advances anomalydetection in space flight operations, potentially enhancing thereliability and safety of future planetary exploration missionsthrough early anomaly detection and proactive maintenance.The model’s ability to capture subtle anomalies, potentiallyindicating early-stage failures, holds promise for the field.